[CostModel][ARM] Increase cost of insert/extract operations

Summary:
This change limits the minimum cost of an insert/extract
element operation to 2 in cases where this would result
in mixing of NEON and VFP code.

Reviewers: rengolin

Subscribers: mssimpso, aemerson, llvm-commits, rengolin

Differential Revision: http://reviews.llvm.org/D12030

llvm-svn: 245225

1 files changed, 12 insertions, 5 deletions

diff --git a/llvm/lib/Target/ARM/ARMTargetTransformInfo.cpp b/llvm/lib/Target/ARM/ARMTargetTransformInfo.cpp
index 38f1821d312..51ece526a6e 100644
--- a/llvm/lib/Target/ARM/ARMTargetTransformInfo.cpp
+++ b/llvm/lib/Target/ARM/ARMTargetTransformInfo.cpp
@@ -255,12 +255,19 @@ int ARMTTIImpl::getVectorInstrCost(unsigned Opcode, Type *ValTy,
       ValTy->getScalarSizeInBits() <= 32)
     return 3;
 
-  // Cross-class copies are expensive on many microarchitectures,
-  // so assume they are expensive by default.
   if ((Opcode == Instruction::InsertElement ||
-       Opcode == Instruction::ExtractElement) &&
-      ValTy->getVectorElementType()->isIntegerTy())
-    return 3;
+       Opcode == Instruction::ExtractElement)) {
+    // Cross-class copies are expensive on many microarchitectures,
+    // so assume they are expensive by default.
+    if (ValTy->getVectorElementType()->isIntegerTy())
+      return 3;
+
+    // Even if it's not a cross class copy, this likely leads to mixing
+    // of NEON and VFP code and should be therefore penalized.
+    if (ValTy->isVectorTy() &&
+        ValTy->getScalarSizeInBits() <= 32)
+      return std::max(BaseT::getVectorInstrCost(Opcode, ValTy, Index), 2U);
+  }
 
   return BaseT::getVectorInstrCost(Opcode, ValTy, Index);
 }